[zeromq-dev] ooc bindings for ØMQ
ph at imatix.com
Thu Jun 17 23:11:46 CEST 2010
On Thu, Jun 17, 2010 at 10:59 PM, Brian Granger <ellisonbg at gmail.com> wrote:
>> Now, if you truly want to define 0MQ as a standard way of communicating across applications / languages / Oss, then yes, you need to standardize the data representation. I personally believe this road leads to madness.
> One of my favorite parts of 0MQ is that it doesn't specify the data rep.
Specifying _the_ data rep would be insane. Specifying _a_ data rep
might be useful. Inevitably we're going to have to do this when we
start to layer on top of 0MQ to do things like security, subscription
upstreaming, reliability, etc. These will (I guess) be layered
protocols (and layer apps) put on top of 0MQ but they will need a
simple fast standardized data rep. As soon as you want to put two
strings into a message header you need some representation.
We can avoid the problem for a while, work around it by using
multipart (as we do for adding the request-reply address) but it seems
to me that we're going to have to bite the bullet sooner or later.
Or... make an external dependency on something like protocol buffers.
Which is nastier IMO.
I was not convinced but writing this email kind of convinced me.
More information about the zeromq-dev